1 .\" Copyright (c) 2003-2008 Joseph Koshy
2 .\" Copyright (c) 2007 The FreeBSD Foundation
3 .\" All rights reserved.
5 .\" Redistribution and use in source and binary forms, with or without
6 .\" modification, are permitted provided that the following conditions
8 .\" 1. Redistributions of source code must retain the above copyright
9 .\" notice, this list of conditions and the following disclaimer.
10 .\" 2. Redistributions in binary form must reproduce the above copyright
11 .\" notice, this list of conditions and the following disclaimer in the
12 .\" documentation and/or other materials provided with the distribution.
14 .\" This software is provided by Joseph Koshy ``as is'' and
15 .\" any express or implied warranties, including, but not limited to, the
16 .\" implied warranties of merchantability and fitness for a particular purpose
17 .\" are disclaimed. in no event shall Joseph Koshy be liable
18 .\" for any direct, indirect, incidental, special, exemplary, or consequential
19 .\" damages (including, but not limited to, procurement of substitute goods
20 .\" or services; loss of use, data, or profits; or business interruption)
21 .\" however caused and on any theory of liability, whether in contract, strict
22 .\" liability, or tort (including negligence or otherwise) arising in any way
23 .\" out of the use of this software, even if advised of the possibility of
33 .Nd "performance measurement with performance monitoring hardware"
43 .Op Fl M Ar mapfilename
45 .Op Fl O Ar logfilename
46 .Op Fl P Ar event-spec
47 .Op Fl R Ar logfilename
48 .Op Fl S Ar event-spec
50 .Op Fl U Ar event-spec
63 .Op Fl o Ar outputfile
64 .Op Fl p Ar event-spec
67 .Op Fl s Ar event-spec
68 .Op Fl t Ar process-spec
69 .Op Fl u Ar event-spec
72 .Op Fl z Ar graphdepth
73 .Op Ar command Op Ar args
77 utility measures system performance using the facilities provided by
82 utility can measure both hardware events seen by the system as a
83 whole, and those seen when a specified set of processes are executing
85 If a specific set of processes is being targeted (for example,
88 option is specified, or if a command line is specified using
90 then measurement occurs till
92 exits, or till all target processes specified by the
94 options exit, or till the
96 utility is interrupted by the user.
97 If a specific set of processes is not targeted for measurement, then
99 will perform system-wide measurements till interrupted by the
102 A given invocation of
104 can mix allocations of system-mode and process-mode PMCs, of both
105 counting and sampling flavors.
106 The values of all counting PMCs are printed in human readable form
107 at regular intervals by
109 The output of sampling PMCs may be configured to go to a log file for
110 subsequent offline analysis, or, at the expense of greater
111 overhead, may be configured to be printed in text form on the fly.
113 Hardware events to measure are specified to
115 using event specifier strings
117 The syntax of these event specifiers is machine dependent and is
121 A process-mode PMC may be configured to be inheritable by the target
122 process' current and future children.
124 The following options are available:
125 .Bl -tag -width indent
127 Toggle between showing cumulative or incremental counts for
128 subsequent counting mode PMCs specified on the command line.
129 The default is to show incremental counts.
131 Create files with per-program samples in the directory named
134 The default is to create these files in the current directory.
136 Toggle showing per-process counts at the time a tracked process
137 exits for subsequent process-mode PMCs specified on the command line.
138 This option is useful for mapping the performance characteristics of a
139 complex pipeline of processes when used in conjunction with the
142 The default is to not to enable per-process tracking.
144 Print calltree (Kcachegrind) information to file
150 this information is sent to the output file specified by the
154 Print callchain information to file
160 this information is sent to the output file specified by the
164 Skip symbol lookup and display address instead.
166 List all event names.
167 .It Fl M Ar mapfilename
168 Write the mapping between executable objects encountered in the event
169 log and the abbreviated pathnames used for
173 If this option is not specified, mapping information is not written.
178 in which case this mapping information is sent to the output
179 file configured by the
183 Toggle capturing callchain information for subsequent sampling PMCs.
184 The default is for sampling PMCs to capture callchain information.
185 .It Fl O Ar logfilename
186 Send logging output to file
191 .Ar hostname Ns : Ns Ar port ,
194 does not start with a
200 will open a network socket to host
207 option is not specified and one of the logging options is requested,
210 will print a textual form of the logged events to the configured
212 .It Fl P Ar event-spec
213 Allocate a process mode sampling PMC measuring hardware events
216 .It Fl R Ar logfilename
217 Perform offline analysis using sampling data in file
219 .It Fl S Ar event-spec
220 Allocate a system mode sampling PMC measuring hardware events
224 Use a top like mode for sampling PMCs. The following hotkeys
225 can be used: 'c+a' switch to accumulative mode, 'c+d' switch
226 to delta mode, 'm' merge PMCs, 'n' change view, 'p' show next
227 PMC, ' ' pause, 'q' quit. calltree only: 'f' cost under threshold
229 .It Fl U Ar event-spec
230 Provide long description of event.
232 Toggle logging the incremental counts seen by the threads of a
233 tracked process each time they are scheduled on a CPU.
234 This is an experimental feature intended to help analyse the
235 dynamic behaviour of processes in the system.
236 It may incur substantial overhead if enabled.
237 The default is for this feature to be disabled.
239 Perform a symbol and file:line lookup for each address in each
240 callgraph and save the output to
244 that only resolves the first symbol in the graph, this resolves
245 every node in the callgraph, or prints out addresses if no
246 lookup information is available.
247 This option requires the
249 option to read in samples that were previously collected and
254 Set the cpus for subsequent system mode PMCs specified on the
259 is a comma separated list of CPU numbers, or the literal
261 denoting all available CPUs.
262 The default is to allocate system mode PMCs on all available
265 Toggle between process mode PMCs measuring events for the target
266 process' current and future children or only measuring events for
268 The default is to measure events for the target process alone.
269 (it has to be passed in the command line prior to
276 Specify that the gprof profile files will use a wide history counter.
277 These files are produced in a format compatible with
279 However, other tools that cannot fully parse a BSD-style
280 gmon header might be unable to correctly parse these files.
281 .It Fl f Ar pluginopt
282 Pass option string to the active plugin.
284 threshold=<float> do not display cost under specified value (Top).
286 skiplink=0|1 replace node with cost under threshold by a dot (Top).
288 Produce profiles in a format compatible with
290 A separate profile file is generated for each executable object
292 Profile files are placed in sub-directories named by their PMC
297 which you can get from
301 .It Fl k Ar kerneldir
302 Set the pathname of the kernel directory to argument
304 This directory specifies where
306 should look for the kernel and its modules.
307 The default is to use the path of the running kernel obtained from the
311 Set system-wide performance measurement duration for
316 may be a fractional value.
318 Print the sampled PCs with the name, the start and ending addresses
319 of the function within they live.
322 argument is mandatory and indicates where the information will be stored.
327 this information is sent to the output file specified by the
330 This option requires the
332 option to read in samples that were previously collected and
337 Set the default sampling rate for subsequent sampling mode
338 PMCs specified on the command line.
339 The default is to configure PMCs to sample the CPU's instruction
340 pointer every 65536 events.
341 .It Fl o Ar outputfile
342 Send counter readings and textual representations of logged data
345 The default is to send output to
347 when collecting live data and to
349 when processing a pre-existing logfile.
350 .It Fl p Ar event-spec
351 Allocate a process mode counting PMC measuring hardware events
357 Set the top of the filesystem hierarchy under which executables
358 are located to argument
362 .It Fl s Ar event-spec
363 Allocate a system mode counting PMC measuring hardware events
366 .It Fl t Ar process-spec
367 Attach process mode PMCs to the processes named by argument
371 may be a non-negative integer denoting a specific process id, or a
372 regular expression for selecting processes based on their command names.
373 .It Fl u Ar event-spec
374 Provide short description of event.
378 Print the values of all counting mode PMCs or sampling mode PMCs
384 may be a fractional value.
385 The default interval is 5 seconds.
386 .It Fl z Ar graphdepth
387 When printing system-wide callgraphs, limit callgraphs to the depth
388 specified by argument
394 is specified, it is executed using
397 To perform system-wide statistical sampling on an AMD Athlon CPU with
398 samples taken every 32768 instruction retirals and data being sampled
402 .Dl "pmcstat -O sample.stat -n 32768 -S k7-retired-instructions"
406 and measure the number of data cache misses suffered
407 by it and its children every 12 seconds on an AMD Athlon, use:
408 .Dl "pmcstat -d -w 12 -p k7-dc-misses firefox"
410 To measure instructions retired for all processes named
413 .Dl "pmcstat -t '^emacs$' -p instructions"
415 To measure instructions retired for processes named
417 for a period of 10 seconds use:
418 .Dl "pmcstat -t '^emacs$' -p instructions sleep 10"
420 To count instruction tlb-misses on CPUs 0 and 2 on a Intel
421 Pentium Pro/Pentium III SMP system use:
422 .Dl "pmcstat -c 0,2 -s p6-itlb-miss"
424 To collect profiling information for a specific process with pid 1234
425 based on instruction cache misses seen by it use:
426 .Dl "pmcstat -P ic-misses -t 1234 -O /tmp/sample.out"
428 To perform system-wide sampling on all configured processors
429 based on processor instructions retired use:
430 .Dl "pmcstat -S instructions -O /tmp/sample.out"
431 If callgraph capture is not desired use:
432 .Dl "pmcstat -N -S instructions -O /tmp/sample.out"
434 To send the generated event log to a remote machine use:
435 .Dl "pmcstat -S instructions -O remotehost:port"
436 On the remote machine, the sample log can be collected using
438 .Dl "nc -l remotehost port > /tmp/sample.out"
442 compatible profiles from a sample file use:
443 .Dl "pmcstat -R /tmp/sample.out -g"
445 To print a system-wide profile with callgraphs to file
448 .Dl "pmcstat -R /tmp/sample.out -G foo.graph"
454 may issue the following diagnostic messages:
456 .It "#callchain/dubious-frames"
457 The number of callchain records that had an
459 value for a return address.
460 .It "#exec handling errors"
463 events in the log file that named executables that could not be
468 events that named ELF executables.
472 events that named executables with unrecognized formats.
474 The total number of samples in the log file.
475 .It "#samples/unclaimed"
476 The number of samples that could not be correlated to a known
477 executable object (i.e., to an executable, shared library, the
478 kernel or the runtime loader).
479 .It "#samples/unknown-object"
480 The number of samples that were associated with an executable
481 with an unrecognized object format.
486 Due to the limitations of the
490 compatible profiles generated by the
492 option do not contain information about calls that cross executable
496 files are also only meaningful for native executables.
509 utility first appeared in
514 .An Joseph Koshy Aq Mt jkoshy@FreeBSD.org
518 utility cannot yet analyse
520 logs generated by non-native architectures.